I have A file which is having data like

input.csv

1;1.30453 103.83589;1.306663 103.83587;1.30 103.5;1.30693 103.83883;1.30709 103.835793;1.30725 103.835757;1.30741 103.83572;1.30753 103.8798

2;1.353 103.83589;1.306663 103.83587;1.30693 103.83583;1.30709 103.835793;1.30725 103.835757;1.30741 103.83572;1.30753 103.835

3;1.30653 103.83589;1.306663 103.83587;1.306797 103.83585;1.3693 

.....

I am trying to convert this file to be in the format (I swap the lat and long values position and group them by a square bracket and seperate them by a comma)

output.csv



1,[103.83589,1.30453],[103.83587,1.306663],[103.5,1.30],[103.83883,1.30693],[103.835793,1.30709],[103.835757,1.30725],[103.83572,1.30741][103.8798,1.30753]

.... and so on.

sed 's/;([0-9.]*) ([0-9.]*)/,[2,1]/g' input.csv > output.csv

can do the job but in 2nd line there is 1.306663 103.83587 1.30693 103.83583 group which has no ; separator. If it is not a misprint script should be modified as

sed 's/[; ]?([0-9.]+) ([0-9.]+)/,[2,1]/g' input.csv > output.csv

which produce

1,[103.83589,1.30453],[103.83587,1.306663],[103.5,1.30],[103.83883,1.30693],[103.835793,1.30709],[103.835757,1.30725],[103.83572,1.30741],[103.8798,1.30753]

2,[103.83589,1.353],[103.83587,1.306663],[103.83583,1.30693],[103.835793,1.30709],[103.835757,1.30725],[103.83572,1.30741],[103.835,1.30753]

3,[103.83589,1.30653],[103.83587,1.306663],[103.83585,1.306797];1.3693

With GNU sed 4.2.2, you'd use a simple sed script, like the following:

sed 's/([0-9.]+) ([0-9.]+)/[2,1]/g;s/;/,/g' < input.csv > output.csv

It needs backslashes for parentheses and + in the match pattern (as well as the numbers in the replacement) to make sed regard them as meta characters.

The script consists of two exhaustive replacements for each line: firstly replacing two number-and-dot sequences separated by space with those sequences in opposite order, separated by comma and enclosed in brackets; secondly replacing semi-colon with comma.

Though, with the dual use of , in the output, it might not be easily loaded as a CSV.

perl -F's+|;' -pale '$" = ",";

   $_ = "$F[0],[" . join("],[", map { "@F[2*$_,2*$_-1]" } 1 .. @F/2) . "]";

' yourfile

Brief

The field separator are set at multiple whitespaces or semicolon. Then

excepting for the 1st field, the rest come in pairs. So keeping aside

($F[0]) we look at 2 fields at a time, flip them, & join all these pairs

with "],[".